For decades, measuring change was a matter of patience. Accumulating observations over years and drawing the line that best describes them. But there is a problem hidden beneath almost every geophysical time series: noise. Not the random noise that cancels itself out, but a more treacherous kind, correlated over time, that distorts the very trend one is trying to extract.
Trajectory estimation is what is at stake when studying rising temperatures, sea level rise linked to climate change, or position shifts caused by the vertical movement of the earth and tectonic plates. These are slow phenomena, measured over years, in which the trend is precisely what matters. And this is where correlated noise becomes dangerous, because it significantly impacts the accuracy of the linear trend estimation. Ignoring it does not solve the problem, it merely hides it inside a line that looks steadier than it should.
Hector exists to correct this. It is a geophysical time series analysis software package that estimates the linear trend and the noise model at the same time, not in separate steps. It uses the maximum likelihood method to determine, in a single process, the slope of the trend and the parameters of the noise that accompanies it. The result is not just a number, it is a number with its uncertainty quantified alongside it.
The tool starts from an honest assumption: the user knows, from the outset, the type of correlated noise present in their observations. From there, Hector allows annual, semi-annual and other periodic signals to be included in the process, and offsets to be estimated at specific epochs. Everything that distorts the reading in a real series becomes part of the model, instead of staying hidden in the margin of error.
Estimating trends has always been the work of science. What changes with a tool like Hector is the honesty of the result. In an age when so much research rests on long, imperfect series, telling the real signal apart from the noise that distorts it is no longer a technical detail. It is the difference between a defensible conclusion and a statistical illusion.